Statistical Evaluation of Seismic Velocity Models of Permafrost

The warming climate in high-latitude permafrost regions is leading to permafrost degradation. Estimating seismic wave velocities in permafrost could help predict the geomechanical properties of permafrost and provide information to plan and design resilient civil infrastructure in cold regions. This paper evaluates the performance of seven models when predicting the seismic wave velocities of permafrost statistically; these models are the time-average, Zimmerman and King, Minshull et al., weighted equation, three-phase, Biot-Gassmann theory modified by Lee (BGTL), and Dou et al. models. The data used in the evaluation are from published laboratory and in situ data, which includes 369 data points for joint P and S wave velocities from nine publications and 943 unfrozen water content data points from 12 publications. The unfrozen water content that is used in these models is determined from a modified Dall'Amico's model that is proposed, which is evaluated against six existing unfrozen water content models based on soil temperature. This paper finds that saturated nonsaline permafrost generally shares similar linear trends between the P and S wave velocities, regardless of soil type, porosity, grain size, and temperature. Fitting all existing data, an empirical linear relationship is derived between the P and S wave velocities. Among the seven models evaluated, the Minshull et al. and BGTL models are the most accurate when predicting the seismic velocities of permafrost.